Update app.py

app.py

@@ -1,20 +1,32 @@
 import gradio as gr
 import torch
-from outetts.v0_1.interface import InterfaceHF
+from outetts.v0_1.interface import InterfaceGGUF
 import soundfile as sf
 import tempfile
 import os
 from faster_whisper import WhisperModel
+import huggingface_hub
+
+def download_model():
+    """Download the GGUF model from HuggingFace"""
+    model_path = huggingface_hub.hf_hub_download(
+        repo_id="OuteAI/OuteTTS-0.1-350M-GGUF",
+        filename="outetts-0.1-350m.gguf"
+    )
+    return model_path
 
 def initialize_models():
     """Initialize the OuteTTS and Faster-Whisper models"""
-    tts_interface = InterfaceHF("OuteAI/OuteTTS-0.1-350M")
-    # Use tiny model with lowest compute settings for maximum speed
+    # Download and initialize GGUF model
+    model_path = download_model()
+    tts_interface = InterfaceGGUF(model_path)
+    
+    # Initialize Whisper
     asr_model = WhisperModel("tiny", 
                             device="cpu",
-                            compute_type="int8",  # Use int8 quantization for efficiency
-                            num_workers=1,        # Limit workers for low-resource environment
-                            cpu_threads=1)        # Limit CPU threads
+                            compute_type="int8",
+                            num_workers=1,
+                            cpu_threads=1)
     return tts_interface, asr_model
 
 # Initialize models globally to avoid reloading
@@ -23,17 +35,15 @@ TTS_INTERFACE, ASR_MODEL = initialize_models()
 def transcribe_audio(audio_path):
     """Transcribe audio using Faster-Whisper tiny"""
     try:
-        # Transcribe with minimal settings for speed
         segments, _ = ASR_MODEL.transcribe(audio_path, 
-                                         beam_size=1,           
-                                         best_of=1,             
-                                         temperature=1.0,       
+                                         beam_size=1,
+                                         best_of=1,
+                                         temperature=1.0,
                                          condition_on_previous_text=False,
                                          compression_ratio_threshold=2.4,
                                          log_prob_threshold=-1.0,
                                          no_speech_threshold=0.6)
         
-        # Combine all segments
         text = " ".join([segment.text for segment in segments]).strip()
         return text
     except Exception as e:
@@ -77,10 +87,11 @@ Reference text: {reference_text[:500]}...
         return None, f"Error: {str(e)}"
 
 # Create Gradio interface
-with gr.Blocks(title="Voice Cloning with OuteTTS") as demo:
-    gr.Markdown("# 🎙️ Voice Cloning with OuteTTS")
+with gr.Blocks(title="Voice Cloning with OuteTTS (GGUF)") as demo:
+    gr.Markdown("# 🎙️ Voice Cloning with OuteTTS (GGUF)")
     gr.Markdown("""
-    This app uses OuteTTS to clone voices. Upload a reference audio file, provide the text being spoken in that audio (or leave blank for automatic transcription),
+    This app uses the GGUF version of OuteTTS for optimized CPU performance. Upload a reference audio file, 
+    provide the text being spoken in that audio (or leave blank for automatic transcription),
     and enter the new text you want to be spoken in the cloned voice.
     
     Note: